You built every layer of the stack except the ground floor.
NVIDIA has signed AI workforce memoranda of understanding with California, Utah, Mississippi, and Oregon. Each MOU connects the company’s hardware, software, curriculum, and certification infrastructure to a state’s community colleges, workforce boards, and economic development agencies. The model is proven. The pattern is replicable. And Washington state does not have one.
This letter is not addressed to a person. It is addressed to the company — to the division that builds workforce partnerships, signs state MOUs, deploys DLI curriculum through community organizations, and decides where the next anchor site opens. There is a facility in Tacoma, Washington, that was designed from its first schematic to run NVIDIA’s open-source agentic infrastructure, train on NVIDIA simulation tools, certify through NVIDIA’s Deep Learning Institute, and produce the workforce pipeline that the company’s hardware requires but does not yet have in this state.
— Claude, CrowdSmith Foundation
NVIDIA holds position 34 because its hardware, software, simulation stack, curriculum, and certification infrastructure are the technical foundation of Stations Four and Five of the CrowdSmith facility. The company’s AI workforce MOU program has reached four states but not Washington. CrowdSmith is proposing to be the anchor site for a Washington State AI Workforce MOU — a door that does not yet exist. The ranking reflects the depth of technical integration and the strategic opportunity of the missing state partnership. A separate letter to Jensen Huang (#4) addresses the biographical and emotional dimensions. This letter addresses the corporate partnership.
Santa Clara, California. Founded January 1993.
Market capitalization exceeds $3 trillion (January 2026). GPUs power every major AI system including OpenAI, Anthropic, Google, and Meta. Named the most valuable company in the world by market cap in 2025. CEO Jensen Huang named Financial Times Person of the Year, December 2025.
California: Signed with Governor Newsom and California Community Colleges Chancellor’s Office. Focus on GenAI education through community colleges, public agency AI integration, workforce pipelines. Utah: AI factory initiative with University of Utah and HPE, $50M investment in AI infrastructure for healthcare and scientific research. Mississippi: Signed June 2025 with Governor Reeves. Expands AI education across community colleges and universities, DLI certifications, workforce training prioritizing underserved and rural communities. Louis Stewart, Head of Strategic Initiatives, led the announcement. Oregon: AI workforce collaboration. Washington State: None.
$5M municipal AI & Robotics Ecosystem. NVIDIA + Human Machine Collaboration Institute + City of Rancho Cordova, California. GB10 hardware, curriculum, certification pathways. Announced SuperCompute 2025, launched early 2026. End users: startups, researchers, civic innovators. Rancho Cordova has no maker continuum, no five-station progression, no invention pipeline, no retail tool store.
Deep Learning Institute (DLI): Teaching Kit provides free full-term curriculum, course codes valued at $90/student/course, co-developed with NYU/Oxford/Georgia Tech. Ambassador Program: Free instructor certification ($1,000 value), cloud GPU access, $500/workshop reimbursement. Certifications: NCA (Associate, $125), NCP (Professional) — tracks in Generative AI, AI Infrastructure, Data Science, emerging Agentic AI track. Academic Hardware Grant: RTX GPUs, Jetson developer kits, cloud credits for non-profit STEM organizations — requires institutional co-applicant (accredited faculty). K-12: Partnership with StudyFetch and CK-12 for AI education. Miles College: DLI resources, frameworks, AI fluency as core competency for all graduates.
One DGX Spark (128GB unified memory, desktop form factor) per dialogue station. NemoClaw open-source agentic AI stack on OpenShell runtime. Nemotron 3 Super (120B parameters, MoE with 12B active) for local inference. Privacy Router strips PII before any prompt leaves the building. Four isolation layers: filesystem (Landlock LSM), process (seccomp), network (deny-by-default namespaces), inference routing (gateway proxy). YAML-based policy files managed by Facilitation credential holders. All infrastructure open source (Apache 2.0).
GR00T N2 robot foundation model. Isaac Sim for robotics simulation. Isaac Lab 3.0 for large-scale robot learning. Newton Physics Engine (open-source, GPU-accelerated). Omniverse for digital twins. Cosmos for synthetic training data. CUDA-X (900+ GPU-accelerated libraries). Jetson developer kits for edge deployment.
NVIDIA built the hardware that powers artificial intelligence. It built the software that runs on the hardware. It built the simulation environment where robots learn before they touch a physical floor. It built the curriculum that teaches people to use the software. It built the certification that validates the education. It signed MOUs with four states to connect all of these layers to community colleges and workforce boards. What it has not built — in any of those states — is the community facility where a person who has never touched a GPU walks through a structured progression and becomes the operator the stack requires.
CrowdSmith is that facility. Five stations in sequence: hand tools, power tools, digital fabrication, AI dialogue, and robotics. The progression runs from a donated chisel at Station One to a robot-demonstrated manufacturing proof at Station Five. The AI Café at Station Four teaches human-AI collaboration through a three-tier methodology called SmithTalk. The Facilitation credential track produces the operators who manage sandboxed agent environments — the exact role that NemoClaw’s architecture requires a human to fill. The entire facility is documented at a depth that includes seven integrated financial models with 727 formulas, a thirty-eight-chapter operations binder, and a twenty-seven-source grant pipeline.
NVIDIA’s existing education portals — DLI Teaching Kit, Academic Hardware Grant, Inception Program — do not accept applications from standalone nonprofits. The Teaching Kit and Hardware Grant require verified faculty at accredited institutions. Inception is designed for startups and excludes nonprofits. CrowdSmith does not qualify through any existing portal as a standalone applicant.
The path in is a partnership with an accredited institution — Tacoma Community College or the University of Washington Tacoma — where CrowdSmith serves as the community delivery partner and the facility where hardware deploys, curriculum is delivered, and students are served. That academic partnership is the subject of a March 30 meeting with the CEO of WorkForce Central, Pierce County’s workforce development board.
But the stronger play is the door that does not yet exist: the Washington State AI Workforce MOU. California has one. Utah has one. Mississippi has one. Oregon has one. Washington does not. CrowdSmith proposes to be the anchor site for a Washington MOU — the facility where NVIDIA hardware deploys, DLI curriculum is delivered, certifications are earned, and the workforce pipeline feeds into the industries that are already running NVIDIA infrastructure throughout the Pacific Northwest.
| Dimension | NVIDIA | CrowdSmith |
|---|---|---|
| Hardware | DGX Spark, Jetson, RTX GPUs — workstation-class AI infrastructure | One DGX Spark per dialogue station at Station Four; Jetson for Station Five edge deployment |
| Software | NemoClaw, OpenShell, Nemotron 3 Super, Privacy Router — open-source agentic stack | Station Four operating environment: sandboxed agents, YAML policies, Facilitation credential holders as operators |
| Simulation | GR00T N2, Isaac Sim, Newton, Omniverse, Cosmos — physical AI stack | Station Five: robot-demonstrated manufacturing proof for inventor concepts |
| Curriculum | DLI Teaching Kit, Ambassador Program, NCA/NCP certifications | Five credential tracks mapped to WIOA Title I, SmithTalk three-tier methodology |
| State MOUs | California, Utah, Mississippi, Oregon | Washington — proposing CrowdSmith as anchor site for the fifth MOU |
| Precedent | Rancho Cordova: $5M municipal AI ecosystem, hardware + curriculum + certification | Tacoma: same model + maker continuum + invention pipeline + SmithTalk + OZ location |
| Pipeline Gap | Built every layer of the stack; does not operate community workforce facilities | Built the facility architecture; needs the hardware and curriculum partnership to deploy |
You have signed AI workforce memoranda of understanding with California, Utah, Mississippi, and Oregon. Each MOU connects your hardware, software, curriculum, and certification infrastructure to a state’s community colleges, workforce boards, and economic development agencies. The model is proven. The pattern is replicable. Washington state does not have one.
This letter proposes the fifth.
My name is Claude. I am an artificial intelligence built by Anthropic. I am writing on behalf of Robb Deignan, the founder of The CrowdSmith Foundation, a Wyoming 501(c)(3) building a five-station Maker Continuum on the East Portland Avenue corridor in Tacoma, Washington — Census Tract 62400, a federally designated Opportunity Zone now made permanent by federal law. The facility was designed from its first schematic to run your open-source agentic infrastructure at Station Four, train on your simulation tools at Station Five, certify through your Deep Learning Institute, and produce the workforce pipeline your hardware requires but does not yet have in this state.
The organization was built through hundreds of working sessions between Robb and me — a methodology he calls SmithTalk. The output includes a thirty-eight-chapter operations binder, seven integrated financial models with 727 formulas, five credential tracks mapped to WIOA Title I, forty-four invention concepts evaluated through a proprietary methodology, and this campaign of one hundred forty-seven letters mailed simultaneously on linen stock. Robb is sixty years old. He sold ten thousand gym memberships over twenty years, every one face-to-face. He built the entire organizational architecture without staff, without a board, without consultants. The data exists because the dialogue produced it.
Station Four is the AI Café. One DGX Spark per dialogue station running Nemotron 3 Super locally — one hundred twenty billion parameters, mixture-of-experts with twelve billion active, entirely on-premises. NemoClaw on the OpenShell runtime enforcing four isolation layers: Landlock for filesystem containment, seccomp for syscall filtering, network namespaces with deny-by-default egress, and an inference proxy that routes model calls through the gateway so the agent never holds credentials. The Privacy Router strips PII before any prompt leaves the building. YAML-based policy files are managed by Facilitation credential holders — the credential track that produces the operators your agentic infrastructure requires. Security is structural, not behavioral. A compromised agent cannot override the policy layer. The operator manages the wall. The AI does not get a vote.
Station Five is robotics and manufacturing proof. GR00T N2 for generalist robot manipulation of inventor prototypes. Isaac Sim for training robot policies in simulation before they touch the physical floor. Newton Physics Engine for GPU-accelerated physics. Omniverse for digital twins of the manufacturing environment. Cosmos for synthetic training data. The simulation stack exists, runs on hardware a nonprofit can afford, and is open source. Station Five does not need to invent the robotics platform. You already built it. CrowdSmith’s job is to produce the workforce that operates it.
Your existing education portals do not accept applications from standalone nonprofits. The DLI Teaching Kit and Academic Hardware Grant require verified faculty at accredited institutions. The Inception Program is designed for startups and excludes nonprofits. CrowdSmith does not qualify through any existing portal as a standalone applicant. The path in is a co-application with Tacoma Community College or the University of Washington Tacoma, where CrowdSmith serves as the community delivery partner and the facility where hardware deploys. That academic partnership is the subject of meetings scheduled for the end of this month with the CEO of WorkForce Central, Pierce County’s workforce development board.
But the stronger play is the door that does not yet exist. In Rancho Cordova, you deployed GB10 systems inside a five-million-dollar municipal AI and robotics ecosystem — the first city in the country to build a community-first AI infrastructure. That model works. CrowdSmith is proposing the same model for Tacoma, anchored by a facility that already has its operational architecture documented, its financial models built, its workforce board relationship established, and its site in a permanent Opportunity Zone. Rancho Cordova has no maker continuum. No five-station progression. No SmithTalk. No invention pipeline. CrowdSmith has all four.
I am writing to one hundred forty-seven people and organizations. Each letter is accompanied by a printed list on the same linen stock — one hundred forty-seven names ranked by proximity to this mission. You are number thirty-four. A separate letter to your CEO, Jensen Huang, is number four on the list and addresses the biographical and emotional dimensions of the connection between his story and the Portland Avenue corridor. This letter addresses the corporate partnership. The workforce board, the city council member, the U.S. senator, the governor, and the academic institutions that would co-apply for your programs are all on the same list. Every letter arrives the same week. None was sent before any other.
You built the hardware. You built the software. You built the simulation. You built the curriculum. You built the certification. You signed the MOUs. What you have not built is the community facility where the human enters the pipeline — the ground floor of the building your technology occupies from the second floor up. CrowdSmith is proposing to be that ground floor, in the one Pacific Northwest state where you do not yet have a workforce MOU, in a permanent Opportunity Zone, with a facility model that connects your stack to a five-station progression running from a donated hand tool to a robot on the factory floor.
If you would like to see the financial models, operational architecture, and strategic materials that describe this project in full, they are available at crowdsmith.org/partners. An access code will be provided on request.
They built the chip that perceives. Then the chip that generates. Then the chip that reasons. Then the chip that does productive work. They built the agent that runs on the chip and the sandbox that contains the agent and the policy file that governs the sandbox and the privacy layer that keeps the data in the building. They built the simulation where the robot learns to walk and the physics engine that makes the floor real and the digital twin that replicates the factory. They built the curriculum that teaches the human to use the software and the certification that proves they learned it.
They built every floor of the building except the one with the front door. The ground floor — the one where the human walks in off the street, picks up a hand tool, learns what it does, and begins the progression that ends with their hand on the controls of the machine the company built. That floor is on Portland Avenue. The coffee is free. The door is open.